DETECTION OF SIGNATURES FROM INTERNAL CONTAMINANT SOURCES USING INTELLIGENT ALGORITHMS

Abstract

Electrical odors and smoke incidents in aviation have become a pressing concern, with over half of the detector activations resulting in false alarms, leading to uncertainties for flight crews. The escalating costs of diversions and growing awareness of associated health risks underscore the need for more reliable detection and discrimination from false alarms. This study harnesses advanced multi-sensor array technologies, intelligent algorithms, and Metal Oxide Sensors (MOS) sensors equipped with AI capabilities to detect and analyze signatures from candidate internal contaminant sources located in the cockpit. Printed circuit boards from avionics, aviation cables of different insulation, and external contaminant sources were put to failure testing to analyze the early fire signatures. These signatures were subsequently assessed using clustering algorithms and multivariate analysis to pinpoint distinct markers. Comprehensive gas analysis and light obscuration measurements further characterized the environment. Experiments were executed at both the University of Maryland and the Federal Aviation Administration (FAA) tech center, replicating diverse conditions, including an altitude simulation of 8000 ft. The focus was on the capability to distinguish between samples during the smoldering phase, leveraging a multivariate approach and gas analysis. The study also incorporated Aspirating Smoke Detection (ASD) to characterize the responses during large-scale testing. The findings pave the way for identifying and integrating innovative technologies, achieving accurate detection of early-stage signatures from internal contaminants during potential aircraft smoke events.